LFS-Aware Surface Reconstruction

Official implementation of LFS-Aware Surface Reconstruction from Unoriented 3D Point Clouds (arxiv)

Dependencies

• Boost
• Eigen3
• QT5
• CGAL5.6.1
• OpenMP

The local feature size (LFS) estimation is available on this repository. We fork from the official CGAL library, and implement a local feature size estimation package. The implementation is here.

To use the LFS package, install the CGAL dependencies and git clone our customized CGAL branch.

git clone https://github.com/bizerfr/cgal.git
cd cgal
git checkout psp-lfs

Then, replace the original CGAL path (CGAL_DIR) with our customized CGAL. We provide two examples.

cd Point_set_processing_3/examples/Point_set_processing_3
mkdir build && cd build
cmake ..
make lfs_example_pointset && make lfs_example_tuple

Try on your own point cloud!

Tips for Using the Toolbox

• When modifying parameters, be sure to re-run the corresponding steps that depend on them. For example, if you change any parameters related to Adaptive Meshing, you must also re-execute the following steps: Estimate_sample_facet_size, Lipschitz_continuity_smoothing_SFS, and Adaptive_meshing.
• Load_Polylines is used to load sharp feature curves, which should be stored in a separate file from the point cloud data.
• max_facet_size, min_facet_size_thickenss_ratio should be carefully tuned. Make sure that max_facet_size is larger than the min_facet_size determined by min_facet_size_thickness_ratio.

Full code will be released after the preprint is accepted.